Ethernet Technology 1 Ethernet Technology 2

• History - Xerox : LAN for open office automation (1970s) 3Mbps on 75 Ohm Coax, coverage 1km area - DEC, Intel, and Xerox Co-proposed “Ethernet v1.0” (9/1980) - D.I.X. Ö Ethernet ver 2.0 (11/1982): 50Ohm Coax, 2.8 km

• ANSI/IEEE 802.3 (ISO 8802.3) - Reformulated Ethernet v2 in 1985 - Drop cable, connector technology, times and fields are redefined - It’s used with Ethernet interchangeably, but less meaningful than Ethernet (?retain the original name) - Supplementary documents (regarding to various medium) + Upgrades are published after Feb.1989 Î IEEE 802.3x Ethernet Technology 1 Ethernet Technology 2

NotationNotation ofof EthernetEthernet TechnologyTechnology

# B X / #

Bit rate in Mbps: Distance in 100m: -1 -10 -100 - 2, -5, -36, -1000 -10000 or Medium: Simplified to: Signaling method: - Baseband 10/100Æ T?/TX or F/FX invented by: Bob Metcalfe and David Boggs in 1970 1GE & 10GE - Broadband 1GEÆ (C/S/L)X or T 10GEÆ(S/L/E)(X/R/W) ൩೭ኬǻ Ethernet Technology 3 Ethernet Technology More on Ethernet Family 4 Bus Topology -- 10Base2/5/36 Explanation of the Abbreviations • Connecting nodes via (tap to) coaxial cable • Transmission and receiving over the same media (line) • Basic configuration (Ex: A 10Base5 segment = a coax) : TRL ~ Transceiver’s cable Length

DBN ~ Distance Between Nodes ӕືႝᢑ୔ࢤȐനߏ 500 ϦЁȑ MSL node NPS ~ Nodes Per Segment DBT Terminator Multiple of 2.5m Coaxial Cable tap Extension ? MES MSL ~ Maximum Segment Length ԏวᏔ Ȑനߏ 50 ϦЁȑ TRL ԏวᏔႝᢑ Transceiver Terminator ಖᆄᏔ MES ~ Maximum Extendable Segments ୔ࢤനӭௗ 100 ঁȑ NPS؂Ȑ

• Important parameters: TRL, DBT, NPS, MSL, and MND.

Ethernet Technology 5 Ethernet Technology 6

Star/Tree Topology -- 10BaseT/F, 1Base5 Transmitting and Receiving Frames

• Connecting nodes via (tap to) UTP / FO A B • Transmission and receiving over separate media (line) • Basic configuration (Ex: one 10BaseT segment)

(ӅӕᡏڮHub Number of ports/hub Bus (ғ All stations are in the same “collision domain” s nt eg me me seg nt Physically a star, Hub - multiple repeater logically a bus

A B

Length (PC/Hub-to-PC/Hub) = a “segment” ; 100m, max Star Ethernet Technology 7 Ethernet Technology 8 IEEE 802.3 (MAC) Frame Format IEEE 802.3 versus Ethernet II 7 1 6 6 2 46 d bytes d 1500 4 ˵̌̇˸̆ LEN Preamble SFD DA SA or LLC PAD FCS Type

z Preamble: (10101010 . . . 10101010) for receiving synchronization z SFD: Start Frame Delimiter (10101011) z DA: Destination Address z SA: Source Address z L/T: Length of LLC-Frame / Protocol type z LLC-Frame: Greater than 46 but up to 1500 bytes (LLC/MAC) z PAD: Padding when LLC-Frame < 46 bytes In practice, however, both frame formats can coexist on the same z FCS: Frame Check Sequence (CRC-32) z MAC-frame size -- from DA to FCS physical coax. This is done by using protocol type numbers (type - Min (18+46) bytes to distinguish from collision field) greater than 1,500 in the Ethernet frame. However, different -Max (18+1500) bytes to prevent dominating bandwidth/channel device drivers are needed to handle each of these formats. Ethernet Technology 9 Ethernet Technology 10

Length/Type field Node’s operation

• Repeater’s operation : Copy what received and sent to all, • Used in two different way: no filtering. Refresh preamble and regenerate incoming frames ~ indicating “length of data field” , not including any padding (by IEEE -- a true standard way) • Node’s (NIC) operation modes (upon sensing a frame): ~ indicating the “type of data” (by DIX) 1. Address filtering - unicast and multicasting • How do we know which one is being interpreted? - selects (if MAC address matched) and copies to node If 1500 t L/T > 0 Ö data size/length Í IEEE frame (802.3) 2. Broadcast If L/T > 1500 Ö protocol type Í Ethernet II / DIX frame - sent to all by letting MAC = 48 1’s - used very often for searching for other stations (Discovery) * The reason for this is rooted in how Ethernet become a standard. 3. Promiscuous Å copies any presented-frames • Fortunately, Ethernets’ hardware never use L/T field. - turn off filtering and copy all frames received It is there strictly for the convenience of the higher-layer protocols. (regardless of Destination’s MAC address) - for diagnostic and network traffic monitoring Ethernet Technology 11 Ethernet Technology 12 Important Terms/Parameters for Ethernets Some Key Parameters • InterFrame gap (IPG) = 96 bit-time Path Delay Value ~ time difference between two ~ the minimum period to be waiting for the end of the nodes, in terms of propagation delay last transmitted frame / packet Å Let stations to sense channel idle • Collision Window (= Maximum PDV)* - the time between when a MAC transmits the first bit of a frame Start of Frame End of Frame and when it detect a collision with any node on the LAN IPG IPG IPG - proportional to network diameter (positively) nop q • Max allowable C.W. = 512 bit times (in specification) time - called Slot time - It determines Min frame size and Max network diameter. • JamSize = 32 bit (For collision enforcement) • Min Frame Size = 512 bits (64 bytes) • AttempLimit = 16 (Max # of retransmissions) • BackoffLimit = 10 (Max # of backoff times) • Network Diameter - the max. cable distance between any two stations • MaxFrameSize = 1518 bytes - Collision domain • MinFrameSize = 64 bytes • AddressSize = 48 bits - the nodes that share an Ethernet using the CSMA/CD rules Ethernet Technology 13 Ethernet Technology 14

CSMA/CD (cont’d) CSMA/CD Transmitting a Frame

The soul of Ethernet and IEEE 802.3 collision ~ Carrier Sense Multiple Access/Collision Detection • Operation: A station with data ready for transmission first listens (senses) to

the channel/medium before its frame transmission: N 1. If the channel Æ Idle Ö transmits frame; otherwise, go to 2. 2. If the channel Æ busy Ö continues to listen until idle, then N Has the IPG passed since transmits frame immediately this node received the last bit of any frames ? (1-persistent CSMA/CD) 3. Senses carrier while transmission. If collision detected

during transmission Collision N Ö transmits jamming signal and ceases its transmission

4. Waits a random time (back-off) and attempt to transmit again too much collision (back to step 1) Ethernet Technology 15 Ethernet Technology 16 CSMA/CD (cont’d) Reveiving/sensing Receiving a Frame the next frame Back-off Algorithm - Collision Recovery

N Receiving • Collision back-off and retransmission in Ethernets completed ? ~ Truncated Binary Exponential Back-off Algorithm (BEBA) Y Let n = number of collisions experienced (n 16) Y d Frame too short (collision) ? k = Min(n,10) Å Truncation 15 retries, 16 attampts in total N then Retry in the r th slot-time , 1 d r d 2k N Address matched ? ~ Back-off time RAND(0,2min(n,10) ) Å Waiting for # of slot-time N Y N More bits CRC received ? correct ? • Disadvantage of BEBA Î unfair Y Y Decapsulating header/trailer Length (if)* Y Ö Last-in-First-out effect: correct ? Submit to - Stations with no or few collisions will have a better chance N higher layer Received to transmit before stations that have waited longer. Error frame Framing error Length error successfully Ethernet Technology 17 Ethernet Technology 18

Backoff (cont’d) • Waiting for: 0 ~ (2k-1) slot-time Õ Binary back-off Example: 10Base5 Medium Specification • Waiting for: 0 ~ (ak + b)-1 slot-time Õ Linear back-off • 10Mbps, Manchester encoding • Probability of successful retransmission ? • Use a 50-ohm(D=10mm) coaxial cable (characteristic impedance = 50 ohm) collision (Impedance is a measure of how much voltage must be slot time = 512 bit-time = 2 W applied to the cable to achieve a certain signal strength) W n o p q • The maximum length of a segment is 500m • The maximum length of a 10Base5 network is 2500m • The distance between any two adjacent taps being a multiple reattempt of 2.5m (Tap); A maximum of 100 taps is allowed Randomly choose From the 2k time slots • Slot time = 51.2 us; IFG = 9.6us Collide: • Packet size: 64 bytes (Min) ~ 1518 bytes (Max) A ~ the first time B ~ the second time Ethernet Technology 19 Ethernet Technology 20 Manchester Encoding for 10Base Ethernet • Basic 10Base2 Ethernet LAN (one segment) Coaxial Cable -- speed = 0.77C

Idle Preamble Data Idle 50: 1 0 1 ... 0 1 0 0 ... 1 0

0V -0.225V // // • Bit interval ̐ࠠௗᓐ -1.825V ಖᆄᏔ (50 ohm terminator) ಒጕႝᢑ୔ࢤ (T-Connector) For 10BaseTx: (Thin Coax Cable) = 1/(10Mbps) = 0.1 us Idle Preamble Data Idle 10Base2 Example 1 0 1 . . . 0 1 1 1 . . 0 0

// // • Thin cable (for 10Base2) is thinner and supports fewer taps over shorter distance than thick cable (for 10Base5) • For 10base2 Ethernet, a transceiver is replaced by a simple 100ns 50ns Î 10Mbps T-connector Ethernet Technology 21 Ethernet Technology 22

10Base2 Example CSMA/CD Collision for the Farest Stations

A B D C

t1 : B and C ready to TX, but B defers upon sensing the carrier from A time t2 : C detects collision and ceases TX

collision

t3 : collision propagates to A and A ceases TX

Ethernet Technology 23 Ethernet Technology CSMA/CD 24 Collision Detection and Late Collision Extensions of Bus (10Base2/5) Networks A B • Two segments (Cable) Ethernet (10Base5 for each segment) Start of A’s frame time Start of B’s frame ӕືႝᢑ୔ࢤȐനߏ 500 ϦЁȑ

B C A jammed

ಃ΋ࢤӕືႝᢑ End of B’s frame End of A’s frame Collision of ԏวᏔႝᢑȐനߏ 50 ϦЁȑ two frames D • An undetected collision ( ૻဦቚ੻Ꮤ (repeater) E I Why ? A B time ಃΒࢤӕືႝᢑ 50m max

F G H

Collision ӕືႝᢑ୔ࢤȐനߏ 500 ϦЁȑ back Ethernet Technology CSMA/CD 25 Ethernet Technology 26

Extending 10Base5 Ethernet with Repeater/Half-repeater Slot time (Collision window) calculation ϡҹ ϡҹ ߻຾ၡ৩ ӣำၡ৩ ᒨۯᒨ ϡҹኧ ϡҹኧ ᏾ᡏۯᒨ ௴୏ۯۓA (max length A ' N = 2,500m) ϡҹ ᛙ 1 D ጓዸᏔ 0.1us 0.1us 5 5 2.0us ୔ࢤ 1 ԏวᏔႝᢑ 5.13ns/ϦЁ 0300ϦЁ 300ϦЁ 3.08us ୔ࢤ 2 B E ԏวᏔȐ໺ଌၡ৩ȑ 0.05us 0.3us 3 3 2.10us ԏวᏔȐௗԏၡ৩ȑ 0.05us 0.6us 3 0 1.95us C F ԏวᏔȐፂናၡ৩ȑ 0 0.9us 0 3 2.70us ૻဦቚ੻Ꮤ (repeater) ӕືႝᢑ 4.33ns/ϦЁ 0 1500ϦЁ 1500ϦЁ 12.99us 2 ъૻဦቚ੻Ꮤ໔ႝᢑ 5.13ns/ϦЁ 0 1000ϦЁ 1000ϦЁ 10.26us G H ႝᢑ០୏Ꮤ 0.1us 0 2 2 0.40us 3 ୔ࢤ 3 500m, max ႝᢑௗԏᏔ 0.1us 0 2 2 0.40us 4 ૻဦቚ੻ᏔȐቚ੻ၡ৩ȑ0.2us 0.4us 2 0 1.20us ૻဦቚ੻ᏔȐፂናၡ৩ȑ0.2us 0.2us 0 2 0.80us ၩݢགෳ 0 0.2us 5 0 1.00us 5 ፂናୀෳ 0 0.2us 0 5 1.00us I ъૻဦቚ੻Ꮤ, half repeater L ૻဦ΢ϲਔ໔ 0 0.1us 3 0 0.30us 51.2uS ъૻဦቚ੻Ꮤ໔ႝᢑ ȐԿ 70% ܭ500ϦЁೀȑ ୔ࢤ 4 ୔ࢤ 5 J Ȑനߏ 1000 ϦЁȑ M ૻဦ΢ϲਔ໔Ȑҗ 50% 0 2.0us 0 3 6.00us Կ 94% ܭ 500ϦЁೀȑ K 6 N ፂናϩപ 0 0.2us 0 1 0.20us ᒨ 46.38usۯૻဦٰӣ໺ሀനߏ Ethernet Technology (Reference: N.-F. Huang, LAN and HSN) 27 Ethernet Technology 28 Approximating the Slot-time 5-4-3-2-1 Design Rule for Bused Ethernet

• Design of an cabled Ethernet network: • (refer to five-segment 10Base5 configuration) - Max no of segments = 5 • Round-trip propagation delay (A to N ) = 46.38 us - Max no of repeaters = 4 • For 10Mbps Ethernet, Data bits that can be transmitted in this - Max no of populated segments = 3 - Min. TWO of five segments function as an inter-link between duration repeater, called IRL. = 46.38 u10-6 u107 = 463.8 bits - All stations are located in the same ONE collision domain • For convenient, • Repeaters ~ extending the length of the network; segment are not “isolated” (collision ??) Let slot time = 51.2 us Ù 512 bits (• No loop allowed ~ limiting one path (of segments and repeater) between any two stations

(It is user’s responsibility to make it loop-free) Ethernet Technology 29 Ethernet Technology 30

5-4 --1 Rule for 10BaseT Ethernet Ethernet Wiring Examples

5 ~ limited to Five segments; 4 ~ limited to Four Hubs; 1 ~ One collision domain

12-port Hub 10Base5 100m Max 12-port Hub (segment) 12-port Hub 12-port Hub A

B 10Base2

500m Max (100m/segment)

Cat. 3 or better UTP RJ-45 10BaseT - restrictions Æ time constraint of Ethernet (512 bit time = 51.2 uS) - if violated Æ packets lost Æ excessive re-sents degrade network performance or create trouble for application Ethernet Technology 31 Ethernet Technology 32 Comparison of various Ethernet Specifications Throughput - Performance of CSMA/CD

ಔ ᄊ • Recall 10BASE5 10BASE 2 1BASE5* 10BASET* 10BROAD36 10BaseFP ୖ ኧ ...... ໺ᒡ൞ϟ ӕືႝᢑ ӕືႝᢑ ค፿ጨԄ ค፿ጨԄ ӕືႝᢑ 850-nm FO ȑ pairۅȑ ᚈ็ጕ Ȑ75 ኻۅȑ Ȑ50 ኻۅȐ50 ኻ Medium ᚈ็ጕ B transmits ୷ᓎ ୷ᓎ ୷ᓎ ୷ᓎ ቨᓎ ୷ᓎ ೌמૻဦೀ౛ Encoding Manchester Manchester Manchester Manchester ȐDPSKȑ Manchester ...... ၗ਑໺ᒡೲ౗ 10 Mbps 10 Mbps 10Mbps Data rate 1 Mbps 10 Mbps 10 Mbps ࢤനεߏࡋ • Consider time on the medium to be organized into slots؂ 500 ϦЁ 185 ϦЁ 500 ϦЁ 100 ϦЁ 1800 ϦЁ 500 ϦЁ (MSL) (slot time = the max time (2W), from the start of transmission, required to detected a collision) ᆛၡനεߏࡋ 925 2500 ϦЁ 1000 (MES) 2500 ϦЁ ϦЁ 500 ϦЁ 3600 ϦЁ ϦЁ (either a collision or no ࢤനӭёೱ؂ 2* 2* 2* transmission in each slot) ௗπբઠኧҞ 100 depends on 2 30 depends on up to 33/coupler W (NPS) hub’s ports hub’s ports Avg.=m πբઠ໔ຯ 2.5 ϦЁ t 0.5 ϦЁ d 1000m * d 1000m * (DBN) ϐ७ኧ d 200m* Star Star Star Topology Bus Bus Bus/Tree Spacing doesn’t exist

Ethernet Technology * “ 1 ” if per-media segment 33 Ethernet Technology 34

Max distance Data rate Maximum Throughput (w/o collision) propagation timedv / dR W Let ; m L ,W d transmission timeLR / vLm R v • Frame in Ethernets 56 8 48 48 16 12,000 max 32 96 ˵˼̇̆ Signal propagation velocity Average frame length Preamble SFD DA SA L/T LLC PAD FCS IPG

TH 11max ? U o * Sometimes, whole thing excluding IPG Frame Payload + 144 bits R aA(1 ) Æ a packet in Layer1 = 12A 13.44a Frame Size Data Payload Max Throughput Efficiency 1 (bytes) (bytes) Frames/sec Bytes/sec (%) (Aemax o| 0.368 asN of .) 1518 1500 812.744* 1219116 97.53 1280 1262 961.538 1213461 97.08 T/E 1024 1006 1197.318 1204502 96.36 512 494 2349.624 1160714 92.86 256* 238 4528.986 1077899 86.23* 128 110 8445.946 929054 74.32 64 46 14880.952* 684524 54.76 64 32 14880.952* 476191 38.10 FS 64 24 14880.952* 357143 28.57 64 16 14880.952* 238095 19.05 64 8 14880.952* 119048 9.52 64 3 14880.952* 44643 3.57 (U) 2W 51.2Ps (512 bit time) ) 812.744 = 10x106/[56+8+ . . . +(1500 x 8) +32+96] * wire-speed traffic comparison How efficiently the LAN is being (N) High-loado ) T = DP x MF used over a given period of time ? Ethernet Technology 35 Ethernet Technology 36 Why Faster Ethernet ? • Applications Driving Network Growth Intranets, Extranet, and tons of Internet applications emerge • IEEE 802.3u, issued in 3/1995 (drafted in late 1994) Multimedia, Video Conferencing • Goal: to provide a low-cost, Ethernet compatible LAN Scientific Modeling, CAD/CAM, Medical Applications operating at 100Mbps while retaining the same Data/Information Warehousing, Document Management wiring systems, MAC method, and frame format Killer applications ? Call for higher bandwidth . . . • Results: 100Base-T series (as shown)

• Ethernet LANs - the Dominant Network Technology IEEE 802.3 (100 Mbps)

. Installed base of over 70 million nodes

. > 80% Marketing share 100Base-X . Low cost of ownership 100Base-T4 . High reliability 100Base-TX (or T2) 100Base-FX

. Scaleable transmission capacity (10 Æ100 Î1000 Î higher ? ) 4-pair Cat. 3/5 UTP . Simple to install and easy to manage 2-pair Cat. 5 UTP or STP 1-pair F.O. Ethernet Technology 37 Ethernet Technology 38

Characteristics of Fast Ethernet Autonegotiation (in 100Base-X) • Data rate = 100 Mbps - Allows PHYs to automatically detect layer-1 and layer-2 features • IEEE 802.3 CSMA/CD frame format (no priority) or Ethernet 2 of the device on the other end by QUERY, and determine the compatible mode of operation ~ called N-WAY Autonegotiation. • (Frame) Collision may encounter Æ still no “delay guarantee” - Query ~ sending and listening a special link integrity pulse • Bandwidth utilization is not guaranteed to be fair which encoding the PHY’ capabilities • Low bandwidth utilization under heavy load (1-persistence) - Allows plug-and-play operation • Suitable for multimedia communications under moderate load - Configuration the links with no users’ intervention • Hub architecture Î Good fault tolerance to links and nodes (no DIP switch, configuration file, or setup screen) (dedicated/independent path for each station) - Priority for operation modes (from the highest to the lowest): • 100BaseT4 1. 100BaseTX or 100BaseFX, FDX mode (Full DupleX) – Use Voice-grade Category 3 UTP or upper 2. 100BaseT4 4. 10BaseT, FDX – Use four pairs of twisted-pair wires 3. 100BaseTX, HDX 5. 10BaseT, HDX • 100BaseX (TX and FX) - Autonegotiation can be disable (IEEE 802.3u) for explicitly – Category 5 UTP, STP, or Fiber optic configuration required Ethernet Technology 39 Ethernet Technology 40 Important Parameters for Fast Ethernet Comparisons of Ethernet and Fast Ethernet

• Max allowable C.W. = 512 bit times = 5.12uS - called Slot time - It determines Min frame size and Max network diameter.

• Network Diameter = 205m (generally) • InterFrame gap (IPG) = 96 bit-time = 0.96uS • JamSize = 32 bit (For collision enforcement) • AttempLimit = 16 (Max # of retransmissions) • BackoffLimit = 10 (Max # of backoff times) = Ethernet’s • MaxFrameSize = 1518 bytes • MinFrameSize = 64 bytes * Topology rules, encoding/signaling method, cost • AddressSize = 48 bits * Network diameter: FE << E (How to overcome the topology limit?) * FE cannot run over coaxial cable

Ethernet Technology 41 Ethernet Technology 42

Two Classes of Fast Ethernet Hubs Difference between Class I and Class II Repeater/Hub • Class-I Hub - Sharing after digitizing , long latency (<0.7uS) • Class I repeater • Class II repeater - T4 and TX/FX can be on the same hub (heterogeneous interface) ~ sharing after digitized ~ sharing in analog form - Can be only ONE class-I hub in a FE network - Can be stackable or even chasis-based to form a SINGLE hub - More flexible to use than Class-II hubs Collision Convert to detection digital data (To continue on the comparison …) Convert to Convert to Convert to • Class-II Hub Collision digital data digital data digital data detection - Sharing before digitizing , short latency (<0.46uS) (short delay provides higher cabling flexibility) Analog Analog Analog Analog Analog Analog - Difficult to build due to tight propagation delay (46 bit time) data data data data data data - Homogeneous interfaces (all TX/FX/T4s) - Use uplink UTP cable when connecting two Class-II Hubs UTPs(nodes) UTPs(nodes) and is limited to TWO-Hub only - stackable, too. Ethernet Technology (An illustrated sample) 43 Ethernet Technology 44 Dual-speed Ethernet Hub Dual-speed Ethernet Hub (cont’d)

SwitchingSwitching circuitcircuit

• The switching circuit does not join 10 and 100 Mbps collision domain • Plan networks containing both 10Mbps (before migration) and 100Mbps together (two repeaters are two separate collision domains (in 916DX)) hosts (actually) two repeaters in one housing devices are automatically • Connection rule depends on what hub/hub stacks it connected to : connected (by Nway detection) to the fastest repeater it can use. - 5-4 rule for 10BaseT Ethernet hubs/stacks, or 10(100) Mbps repeater retransmits Ethernet (Fast Ethernet) - 205m network diameter (of a collision domain) using two Class II Fast transmission to all other ports operating at the same speed Ethernet hubs/hub stacks Ethernet Technology 45 Ethernet Technology Continue . . . 46

Example - Hub)໣ጕᏔ* Stacking – Treats Hubs as a Whole

Daisy Chain

Out Stacking Cable (flat cable)

X ~ switching between 10 and 100 circuitry

Ethernet Technology 47 Ethernet Technology 48 A few things to remember • Hub stacking (back plane) •MDI-X ~ Medium Dependent Interface, cross-wired

•MDI-II ~ Medium Dependent Interface, straight-through

•1X Æ Uplink Ö connected together (inside) Don’t use both the port 1X jack and the Uplink jack at the same time.

•Use uplink method to connect different brands and types of hubs.

Using daisy-chain flat-cable • Daisy chain IN port and OUT port ~ for creating stack. • If hubs are stacked, do not link them using Uplink again.

- Use Daisy-chain connection to stack subs • Stacked hubs does not count as an additional repeater like an uplinked - Do use the same brand and stackable hubs in stacking hub does. They are treated as ONE repeater in the network diameter. - DC cables: D-sub 50 pin (SCSI II) and D-sub 25 pin However, since the stacking cable still introduces a little delay, the * Cable conversion (D-sub 50 to D-sub 25) is also counted as ONE stack MAX # of stacked hubs are limited. Generally, 5 to 7 ‘s. Ethernet Technology 49 Ethernet Technology 50

100BaseTX Fast Ethernet (cont’d) Basic Configuration - One Class I/II Hub + stacking Expanded Configuration

(all 10Mbps devices)

Ethernet 10BaseT-500m Max

MDI-II 100m/segment

Two Class II Hub/Hub stack - Network diameter = 205m, Max. (100+5+100) Fast Ethernet Straight Cross - If the “max. station-hub” <100m, the 100Mbps, 205m Max DTEthrough DCE DTE DTE wired “hub-hub” distance can then > 5m. (ኧᏵಖᆄ೛ഢ) UTP (ኧᏵጕၡ೛ഢ) UTP TIA/EIA 568 A/B standard Ethernet Technology 51 Ethernet Technology 52 General Model of a Switch Crossbar - An Example of switching Fabric

Control N x N N ports FE Switch Subsystem(s) Ù

1 11IPC OPC .

. 1 . . Output conflict Arbiter 2 Arbiter inbound outbound 2 . N IPC OPC N inputs . Buffer/Memory

Switching Fabric .

. Blocking - Layer 2 device (operated at frame-basis, forwarding based on the MAC address) N (Nonblocking or - Exchange information based upon MAC address (learned & stored internally) N 1 2 . . . N Rearrangeable

- Example: a 16x16 switch or a 16-Port switch Input output Switch) - Three operation modes: (viewpoints: latency, error forwarding, buffers, etc) queues (Buffer) Store-and-forward switching, Cut-through switching, and 1 2 . . . N modified/Runt-free/Interim cut-through switching outputs (An illustration) Ethernet Technology 53 Ethernet Technology 54

Hub (໣ጕᏔ) versus Switch (ҬඤᏔ) Switch’sSwitch’s BasicBasic OperationOperation

• Layer-1 device (hub) • Layer-2 device (switch) Learning • Multiple repeater • Multiport bridge • TX/RX ONE pair at a time • TX/RX Multiple pairs at a time • Connect multiple Ethernet • Multiple Ethernet segments operate segments into ONE network separately but reachable (1 collision domain as a whole) (1 collision domain per port) • Collision domain in hubed ntwks • Broadcast domain in switched ntwks Forwarding

MAC address table (basic) : MAC address Port number 0080c8-234568 2

Broadcast domain

ᐱ൧ך΋Γளၰ--୤ ન 55 Ethernet Technology A switched network = a broadcast domain. 56ךՉך--Ethernet Technology ᚈᚈჹჹ ExampleExample -- Switch Switch))ҬҬඤඤᏔᏔ** Design of Switched Ethernet

- Extending Fast Ethernets with Switches to overcome the topology limitation 100m Max FE 100Mbps Hub 100m Max FE 100Mbps Hub FE Switch FE Switch A 100m (FDX - called point-to-point, B w/IPG only, w/o CSMA/CD)

Max forwarding rate 500m or more

Aggregate forwarding rate Between A and B

“oversubscribed” - dedicated bandwidth 100Mbps per port/node

Ethernet Technology 57 Ethernet Technology 58

i Started 11/1995; Standardized 2/1998 Ö IEEE 802.3z Basically . . . GE is i Provides 10 times the performance of Fast Ethernet a Ethernet, i Complete interoperability with Ethernet and Fast Ethernet only - Retain the investment in installed base of network infrastructure FASTER - Leverages on the network management environment i Conforms to the Ethernet standard - Preserve Frame format and frame sizes - CSMA/CD access method & BEBA i Provide a simple forwarding mechanism between 10, 100, and i But, two new supports: Carrier extension and frame burst 1000 Mbps (easy for migration) - No fragmentation, encapsulation, and translation of frames * Reference: http://www.gigabit-ethernet.org Ethernet Technology 59 Ethernet Technology 60 Gigabit Ethernet Communication Architecture 802.3z and 802.3ab Physical Media

Media Access Control (MAC) Full-Duplex and/or Half-Duplex 5um Singlemode 1000BaseLX ~1300 nm Gigabit “Media Independent Interface” { 50um Multimode 62.5um Multimode MAC Fiber channel (1996/6) 1000BaseT UTP (1997/3) Encoder/Decoder Encoding/Decoding 1000BaseSX 50um Multimode PHY ~850 nm { 62.5um Multimode 1300nm 780nm Copper Twisted Pair LWL optics SWL optics wire Transceiver 1000BaesT (802.3ab) { 4 pr Cat 5 UTP

50 µm or 62.5 µm Single-Mode Fiber Balance Twisted Pair Cabling Multi-Mode Fiber 1000BaseCX (9um) 2 km - 10 km (200 m to 550 m) STP 100 m (Cat-5 UTP) Copper C Balanced Shielded { 2-pair STP Cable // Full-Duplex Links Half-Duplex Repeater 25m 100m 260m 440m 550m 3 km or more* 802.3z physical layer 802.3ab physical layer Machine Wiring Building Campus/Business (supports 200 m to 10 km link distances) (supports 200 m network diameters) Room Closet Backbones Backbone

Fiber Channel Technology LWL - Long-Wave Laser; SWL - Short-Wave Laser

Ethernet Technology 61 Ethernet Technology 62

Important Parameters for Gigabit Ethernet Carrier Extension • Remedy (to overcome the discrepancy between MinFrame and slot-time) • Max allowable C.W. = 4096 bit times = 4.096uS ?? - Extend the CSMA/CD slot-time Ö Carrier extension - in order not to restrict the network diameter to 25m only (recall: slot-time = the time unit for Ethernet MAC to handle collisions) and keep the MinFrame = 512 bits for uniform frame structure by adding bits to the frame until the frame meets the minimum slot-time required • InterFrame gap (IPG) = 96 bit-time = 0.096uS Ö In this way, the smaller packet sizes (less than 512 bytes) can • JamSize = 32 bit (For collision enforcement) coincide with the minimum slot-time and allow seamless • AttempLimit = 16 (Max # of retransmissions) operation with current Ethernet CSMA/CD. • BackoffLimit = 10 (Max # of backoff times) Carrier extension : = Ethernet 0 d CE d 448 bytes • MaxFrameSize = 1518 bytes added after FCS • MinFrameSize = 64 bytes collision • AddressSize = 48 bits Station A Short pkt Extended frame • A calculation of GE network delays Ö 25m network diameter d 512bytes 512 bytes • Solution: increase slot-time Station B Ethernet Technology 63 Ethernet Technology 64 Frame Bursting DesigningDesigning andand PlanningPlanning aa Ethernet-basedEthernet-based LANLAN • Goal : Step 1: Placing nodes and determining the speed for nodes: fixed, - Compensating carrier extension scheme to improve utilization dual(or auto-negotiable) • Method : Step 2: Clarifying/Planning the collision domains according to the - adopts pipelining scheme but no change on the MAC interface different working population, geographical distribution, and (transmit and receive ONE frame at a time) traffic/performance consideration (requirement and balance) - transmits multiple frames/packets in a burst )ૻᛈ* Step 3: Equiping the Devices (E/FE, Hub/Switch) and build up the separated by extension carriers infrastructure • Carrier extension must be applied to the first frame so that the Step 4: Determining the Cabling (Fiber or UTP, wires) according to collision will only affect the first frame of a burst Õ Txer and the node distribution and future expansion Rxer are considering “ one frame at a time” as usual Step 5: Operating and testing performance (Basiline) and then modifies somehow if necessary • Burst size would vary between 1518 byte frame (12144 bits) Note: Be provide with high-bandwidth link for server access. to a maximum of 8192-byte (65,536 bits, 5/1997 by IEEE Be design with high-bandwidth for the LAN backbone links. 802.3z) * high-bandwidth Æ 100/1000 Mbps Hub/Switch links Ethernet Technology 65 Ethernet Technology 66

Gigabit Ethernet Migration • Upgrading Switch to Switch • Upgrading Switch to Server Links Connections - to obtain high-speed access to - to obtain 1000Mbps pipes applications and file systems between 100/100 switches (to support a greater number of both switched and shared fast Ethernet segments)

Server farm

Ethernet Technology 67 Ethernet Technology 68 ProblemsProblems inin SwitchedSwitched NetworksNetworks • Upgrading a Switched Fast Ethernet Backbone • Broadcast Storming FE Switch 1 - by aggregating fast Ethernet FE 100Mbps Hub 2 FE 100Mbps Hub 1 switches with a Gigabit Ethernet (coming back) switch or repeater (to support more nodes and A bandwidth per segment) FE Switch 2 Station A send a broadcast frame B (w/ 48 1’s in DA)

• The broadcast sending by A is back to FE Hub1 ~ called Frame Cloning Caused by a looping (i.e., a graph) (ᘉ஭ᐋᄽᆉݤ) • Solution ~ by Spanning-tree algorithm to prune a looped network automatically into a loop-free network (from a graph to a tree) Ethernet Technology 69 Ethernet Technology 70

(Binary math. representation) Cyclic Redundancy Check (CRC) • Effectively, • Idea of CRC ˧ˠ˙ ˅ ́ Generates F by P(x) or a number, P shift M left n-bit; padded in with n zeros • In case of no error: k-bit message, M n-bit FCS, F Transmitter’s action We want T to be divisible by P (pattern of n+1 bits) : frame transmitted, T ˅ ́ ˠ ˧ˠ˥ ˅ ́ transmitted Î ˣ ˤ ˥ Q: Does this R satisfy the conditionˣ that T/P have no remainder ? Receiver’s action Received frame, Tr A: T R R Q Q exactly divided by P divided by the pre- P P P (n+1 bits) determined number, P Really? Å true, if modulo-2 arithmetic is applied.

n If remainder = 0 No error Thus, taking FCS = Remainder of “ 2 M divided by P ”. Ethernet Technology 71 Ethernet Technology 72 Modulo-2 Arithmetic Example: ~ binary operation without carriers/borrower (i.e., the “exclusive” logic) 1111 11001 ( + 1010 u 11 1, 0101 1001011

Example: (CRC calculation )

M = 1010001101 P = 110101 LSB FCS R = 01110 ÖÖ T = 101000110101110

If Tr = T+E (may be) & Tr/P = 0 ÖÖ E = 0, no error

Ethernet Technology 73 Ethernet Technology 74

Polynomial representation • Notations

• ~ k-bit block data k-terms of xi M(x) = Message/data block polynomial ( P(x) = Generating polynomial (check-bit generator) - Coefficients of F(x) the bits in the binary data (known to the Txer and Rxer) or G(x) - Examples: n = the deree/order of P(x) R(x) = Remainder polynomial 1. ‘101’ x2+1 = F(x) (‘0’ Æ missing term) • CRC encoding process 2. F(x) = (x2+1)(x3+x+1) = x5+x2+x+1 ‘100111’ 1. Appends n’s ‘0’ to M(x) Ö xnM(x) 2. Divide xnM(x) by P(x) Ö ̋ˠ́ ʻʼ ̋ ˤ ̋ ˣ ʻʼʻʼ ̋ ˥ ̋ ʻʼ * Remark: 3. Transmit the encoded message - If an F(x) contains (x+1) as a factor, ́ then F(1) = 0. (Why ?) ˧̋ʻʼ ̋ˠ̋{ ʻʼ ˥̋ ʻʼ concatenate here Checksum (FCS) Ethernet Technology 75 Ethernet Technology 76 CRC Error Detection SomeSome InternationalInternational standardstandard ofof P(x)P(x) oror G(x)G(x)

12 11 3 2 CRC12: P(x) x x x x 1 • Error detection capability of CRC CRC16: P(x) x16 x15 x2 1 - All single-bit error - All double-bit errors, as long as P(x) has at least three 1’s CRCCCITT: P(x) x16 + x12 + x5 +1 - Any odd number of errors, as long as P(x) contains (x+1) 32 26 23 22 16 12 CRC-32: P(x) x x x x x x - Any burst error with burst length < length of P(x) x11 x10 x8 + x7 x5 x4 (i.e., the length of FCS) x2 x1 - Most larger burst error (a small set of bits near a single location)

CRC8: P ( x ) x82 x x 1 ( in ATM cell ) • CRC does not covers preamble and SFD Ethernet Technology 77 Ethernet Technology 78

• Implementation of CRC by digital logic • Transmitter’s action: (by XOR gates and 1-bit shift registers (SR)) - Reset all SR = 0 and shift in message one-by-one, - # of registers = n = the length of the FCS - # of XOR gates up to n MSB first - Checksum (FCS) is left in SR after the last data - the presence (absence) of an XOR bit is transmitted Ù the presence (absence) of a term in P(x) - Shift FCS to channel - All SRs are clocked simultaneously and thus shift along SRs

• General CRC implementation architecture for P(x) • Receiver’s action: ́ ˄ ˅ - Implementing the same H/W as the Txer does ˣ̋ʻʼ ̋ ˴ ̋ ˴̋˄ ˴̋ ˅ ˄ ˄ Input bits ́ ́ XOR SR - Rxed data is shifted to the input end and resulted in R in SR, then R is coming on in - If SR = 0 ÖÖ no error a an-1 an-2 a2 1 Multiplier -If SR z 0 ÖÖ erred

With a = 1, and a = 1 Ethernet Technology 0 n 79 Ethernet Technology 80 • Example Message M = 1010001101 Ù M(x) = x9 + x7 + x3 + x2 + 1 4B/5B-NRZI Data Encoding for 100Base-FX Divisor P = 110101 Ù P(x) = x5 + x4 + x2 + 1 • Use “bit encoding”, instead of clock encoding which is not

c4 c4 c4 appropriate for high speed (like 100Mbps transmission)

gate • Encoding rules: - 4-bit data are encoded into a symbol with 5 code bits (only 16 4-bit data block out of 32 cases are selected) 1 clock - select the code group with at least two transitions presented - no more than three ‘0’s are allowed across one or more Message to be sent code groups - encode 5-bit code groups using NRZI (differential encoding is more reliable in detecting a

Five zeros added transition in the presence of noise and distortion than in comparing a value to a threshold) • Coding efficiency is raised to 80% MSB first ~ 100 Mbps (data rate) is achieved with 125 Mbaud rate

Ethernet Technology 81 Ethernet Technology 82

MLT-3 Encoding for 100BaseTX Fast Ethernet

• 4B/5B-NRZI is effective over Fiber, but in a twisted pair . . . Ö the signal energy is concentrated and thus may produce undesirable emission from the wire Overcome Ö Use MIL-3 encoding scheme to concentrate most of the Txed-signal energy below 30MHz

• MIL-3 encoding steps (rooted from 4B/5B NRZI) : 1. Convert NRZI back to NRZ 2. Scramble bit stream to produce a more uniform spectrum

4B/5B-groups NRZI code 3. Follows the encoding procedure as shown in the state diagram (Scramble – No reduction in data rate) • Keynote: Every time there is an input of 1, there is a transition. • Drawback: lacks DC balance The Occurrences of +V and –V alternate.

Ethernet Technology 83 Ethernet Technology 84 MIL-3 Encoder’s State Diagram

THE END

• Example (assumed not being scrambled)

Ethernet Technology 85 Ethernet Technology 86